The measurement of receptor tyrosine kinases (RTK) by standard immunohistochemistry (IHC) has several limitations including the lack of sub-cellular compartmental measurements which can be indicative of intracellular signaling. Simultaneous and automated measurement of membranous, cytoplasmic and nuclear levels of RTKs would offer a superior approach to traditional IHC and may be useful for the elusive molecular stratification of various cancers. The RTK hepatocyte growth factor receptor (Met) is believed to play a role in the progression of colon cancer, however the association of Met expression and survival has not been well established for colon cancer. We performed fluorescent-based immunohistochemistry for Met in 583 colon cancer patients in a tissue microarray format. Using novel curvature and intensity based image analysis, the membrane, nuclear and cytoplasm were segmented, and the epithelial tissue was automatically separated from the stromal tissue. Probability distributions of Met within each compartment were determined, and an automated scoring algorithm was generated. An optimal cut point was selected using 500-fold cross validation of a training and test data set. In univariate Cox analysis, high cytoplasmic relative to membranous values were a significant predictor of survival in stage I (hazard ratio (HR) 0.16; p=0.006) and in stage II patients (HR 0.34 p = 0.0005). Similar results were found with multivariate analysis. The value of Met in the membrane alone was not a significant predictor of outcome in all patients or within stage. This is the first report to show sub-cellular distribution of Met is a predictive molecular biomarker and our data support the use of automated algorithms for the molecular stratification of early stage colon cancer. Further, these algorithms have general applicability to RTKs that are known to have complicated intracellular signaling patterns.
- American Association for Cancer Research